A Leaf Disease Detection Using Machine Learning and Deep

Learning: Comparative Study

Mooad Al-shalout

, Mohamed Elleuch

and Ali Douik

Computer Science Department ISITCom, University of Sousse, Sousse, Tunisia

National School of Computer Science (ENSI), University of Manouba, Tunisia

National Engineering School of Sousse NOCCS-ENISO Lab, University of Sousse, Sousse, Tunisia

Keywords: Detecting Diseases, Corn, Potato, Tomato, VGG19, SVM, SIFT, Gabor.

Abstract: This study aims to provide innovative methods and additional suggestions for detecting plant diseases using

deep learning techniques. The study focused on identifying diseases affecting major daily consumed plants,

such as tomatoes, corn, and potatoes. The detected diseases included rust, early and late spots, mildew, and

bacterial spots. The study relied on machine learning and deep learning algorithms, such as Support Vector

Machine and VGG19 algorithm, to detect plant diseases. SIFT and Gabor filters were also incorporated into

the work and tested using SVM algorithm. The study reached highly accurate results, as the accuracy rate

reached 98% using SVM, and 97% using VGG19 algorithm, which are satisfactory results compared to

previous studies, confirming the effectiveness of the methods used in detecting plant diseases.

1 INTRODUCTION

Agricultural crops are among the most important

basic pillars of human life, which they depend on

completely in their lives. For thousands of years,

humans have paid attention to agricultural crops,

especially in some developing countries that depend

for their economic components on agriculture

because it is the basic resource in human life, and it

also provides a large portion of Work for some

people.

Sometimes, agricultural crops are afflicted by

diseases, which can be a significant cause of their

complete or partial destruction. Any disease affecting

these plants negatively impacts their quality, either in

terms of health or economic aspects, leading to a

decrease in their value. Farmers often incur

substantial losses, resorting at times to agricultural

experts and pesticides to combat diseases. This is

where artificial intelligence comes into play in

diagnosing diseases affecting medicinal plants. It has

achieved remarkable success in diagnosing and

distinguishing certain diseases, attempting to limit the

spread of diseases among other agricultural crops,

https://orcid.org/0009-0000-1913-658X

https://orcid.org/0000-0003-4702-7692

https://orcid.org/0000-0002-0178-501X

and improving the agricultural and economic sectors

simultaneously by reducing the cost of diagnosing

plant diseases. A large number of machine learning

algorithms have been employed to produce better

crops, training on a wide range of data related to the

agricultural sector to mitigate diseases (Goralski &

Tan, 2020).There are many studies that have proven

its efficiency in detecting plant diseases and working

on them seriously and extensively (Shruthi,

Nagaveni, & Raghavendra, 2019). In this study, we

tried to discover diseases that affect corn, tomatoes,

and potatoes using CNN deep learning algorithms.

These particular plants were selected due to their

status as fundamental crops essential for people's

daily sustenance, ones they frequently find

indispensable (Mohanty, Hughes, & Salathé, 2016).

Research was conducted on the identification of plant

diseases employing both the SVM algorithm and the

VGG19 algorithm.

Each of these attempts led to a high score in some

type of plant, with accuracy reaching approximately

98% in some algorithms.

In this paper, we looked at three main types of

plants that affects human life, as we mentioned

940

Al-shalout, M., Elleuch, M. and Douik, A.

A Leaf Disease Detection Using Machine Learning and Deep Learning: Comparative Study.

DOI: 10.5220/0013240400003890

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2025) - Volume 3, pages 940-947

ISBN: 978-989-758-737-5; ISSN: 2184-433X

previously, which are corn, potatoes, and tomatoes.

There were a number of nematode diseases that affect

these plants, including common rust, leaf spot, and

northern leaf blight in corn plants, and then early

blight and late blight in potatoes, in addition to the

bacterial spot that affects tomatoes and Spectura

leaves, which are among the targeted diseases in

tomato plants.

Work was done on deep learning algorithm,

compared with traditional machine learning methods

such as SVM. These algorithms were chosen because

they achieved good results in detecting and

classifying diseases in many studies (Arora &

Agrawal, 2020). Furthermore, it can be used in many

areas including image classification, and object

detection (Song et al., 2019).

The structure of the paper was as follows: In the

second section, we provided an overview of prior

studies, discussing various research endeavors and

advancements in the scientific field. Following that,

the third section detailed the methodology employed

in conducting our study, transitioning to the

subsequent stage. Here, we presented and analyzed

the results we obtained. The research work concluded

by summarizing our findings and outlining the

intended objectives for future work.

2 RELATED WORKS

Tomato an individual consumes approximately 42

kilograms, especially in North America, and in order

to preserve that plant efforts are made to preserve it

(Albawi, Mohammed, & Al-Zawi). Artificial

intelligence has been used to discover potential

diseases on tomatoes and use some artificial

intelligence applications and algorithms for to an

early detection of those diseases that affect those

plants and classify the condition if the disease is

found or not (Laranjeira et al., 2022).

In (Natarajan, Babitha, & Kumar, 2020),

researchers worked on developing techniques used in

deep learning to detect diseases in a number of plants,

including tomatoes. The most common diseases in

that plant were bacterial spot, leaf curl, bacterial

spots, and early and late end blights of that plant. A

number of techniques were adopted in deep detection

of the plant. Including: Single Shot Detectors (SSD),

VGG, and AlexNet. In addition to the ResNet

algorithm for detecting diseases that affect plants. In

that study, a small number of real images containing

a number of diseases were worked on, and they were

detected in a number of early, intermediate, and final

stages of the disease. The results in that case showed

that the accuracy rate reached 95.71%.

In (Shijie, Peiyi, & Siping, 2017), the researcher

worked on developing a CNN model with transfer

learning algorithms in the VGG16 algorithm to detect

a number of diseases related to plants, such as spider

mite, gray spot, mosaic viruses, targeted bacterial

spots, and leaf spot. A healthy leaf is considered

healthy disease, but there is no injury. A number of

real photographs (7040) were used in this study. The

researcher extracted features from the original images

using VGG16 and compiled them into the Support

Vector Machine algorithm to classify them to

determine the disease and its type. The average

accuracy obtained was 89%, and the deep learning

framework Keras/TensorFlow was used in that study.

Furthermore, in (Arora & Agrawal, 2020)

researchers worked on proposing a new approach to

classify corn leaf diseases through the application of

a number of algorithms, such as Deep Forest. They

used something new to discover three diseases, which

are: leaf spot and rust disease common in plants. In

addition to leaf spot disease, work has been done on

a small dataset consisting of only 400 images, and

these studies have shown good results. The accuracy

in that study in describing and identifying the disease

in corn plants reached 96.25%, while in the algorithm

LeNet5 reached 83.46% accuracy, and finally the

CNN algorithm reached 91.25% accuracy. From

there, the researcher arrived at the approach he

proposed that is capable of competing with traditional

deep learning methods and is a good alternative to

image-based applications.

In (Al-Shalout, Elleuch, & Douik, 2023), the

study employed several algorithms, including

VGG16, VGG19, and CNN utilizing around 25000

images. Among these algorithms, VGG19

demonstrated superior performance, achieving a

remarkable accuracy rate of 95%. The CNN

algorithm also yielded promising results, with an

accuracy rate of 90%, while the accuracy rate for the

VGG16 algorithm reached 86%.

In (Reis & Turk, 2024), a novel approach to plant

disease classification is introduced, utilizing the

Integrated Deep Learning (IDLF) and Ensemble

Learning (EL) framework. This methodology

integrates pre-trained deep neural networks,

including the ImageNet-based model, with 13 distinct

deep learning architectures (DLA), comprising

models trained from scratch and hybrid variations.

Various image quality enhancement techniques, such

as hypercolumn, contrast stretching, and Contrast

Limited Adaptive Histogram Equalization (CLAHE),

were applied. The primary objective is to attain robust

A Leaf Disease Detection Using Machine Learning and Deep Learning: Comparative Study

941

classification performance. In experimental trials, the

RegNetY080 model trained from scratch achieved an

accuracy of 97.64% on the original dataset, which

increased to 98.33% with CLAHE optimization.

In (Alzahrani & Alsaade, 2023), the pressing issue

of early detection of tomato leaf diseases, vital for

sustaining crop quality and yield, was tackled.

Utilizing computer vision and advanced artificial

intelligence, the study utilizes three deep learning

models - DenseNet169, ResNet50V2, and Vision

Transformer (ViT) - to classify tomato diseases.

3 METHODOLOGY

In this section, we present the proposed processes and

methodologies that we worked on (See Figure 1).

Figure 1: Proposed Methodology.

3.1 Phase One: Data Collection and

Data Pre-Processing

3.1.1 Data Collection

The data utilized in our research was sourced from the

Kaggle dataset, comprising 25272 images. These

images encompass various plant diseases, with a

specific focus on three types: tomatoes, corn, and

potatoes, as mentioned earlier. The targeted images

represent authentic plant leaves, featuring both

healthy and diseased specimens within each category.

The observable symptoms on these leaves encompass

leaf spot diseases, bacterial spots, and target spots on

tomato plants, as depicted in Figure 2(a). On corn

plants, the symptoms include leaf blight, common

rust, and leaf spot, as shown in the respective figure

2(b). The figure also illustrates advanced and late

blight, along with bacterial spots on potato plants in

Figure 2(c). The images given in figure 2 illustrate

real pictures of both diseased and healthy plant

leaves.

(a) (b) (c)

Figure 2: (a) Tomato Disease, (b) Potato Disease and (c)

Corn Disease.

3.1.2 Data Pre-Processing

Central processing is one of the key procedures

necessary for extracting data from images. The data

extraction process is defined as a data extraction

method that works to convert raw data into a format

for the purpose of determining the data you are

working with. Before initiating the data extraction

process in images, it is crucial to perform a significant

pre-processing step, primarily involving the central

processing unit. This step is essential because the data

often tends to be perplexing or ambiguous, lacking

accurate and meaningful values necessary for

training, extraction, and obtaining reliable results. To

enhance the quality of the data and achieve optimal

outcomes, thorough cleaning and processing of the

images are carried out prior to the commencement of

the work, as emphasized in (Lazzeri, Bruno, Nijhof,

Giorgetti, & Castoldi, 2015).

Our research involves a three-stage data

processing approach: Labels and image unification in

addition to data normalization.

Labels: This process involves transforming labels

and images into digital representations, enabling easy

comprehension and interpretation by the program. It

facilitates instructing the machine in reading, defining

the utilized control, and managing digital

components.

Table 1: Label encoding.

Tree Label Disease

Corn

(4 Classes)

0 Vercos

ora leaf s

1 Common_rust

2 Northern Leaf Blight

3 Corn health

Potato

(3 Classes)

0Earl

Bli

1 Late Bli

2 Health

Tomato

(4 Classes)

0 Bacterial Spot

1 Health

2Se

toria Leaf S

3Tar

et S

ICAART 2025 - 17th International Conference on Agents and Artiﬁcial Intelligence

942

This constitutes a fundamental step in handling

structured data, serving a supervisory role (Singh &

Singh, 2020). Table 1 displays the encoding present

in the specified dataset.

Data Normalization: It is the process of

converting image pixel values into a more common

or familiar meaning. In this process, image data pixels

(intensity) are projected onto a specific scale or the

data is rescaled (usually (0,1) or (-1,1)). This process

is used when the dataset contains many image formats

and only one algorithm will be applied to it (Reis &

Turk, 2024).

Image Standardization: defined as the process

of controlling the image and its remainder (height

with width), that is, controlling the pixels of the

image, whose goal is to improve quality, standardize

measurements, and maintain consistency for all

images.

Image Data Generation from the Kera's Library

provides a sample for each image and data set, which

is obtained from the average and standard deviation

statistics necessary to standardize the values in the

images, and is done through the individual pixel

values in each image or the groups as a whole

(Weinberger, Seroussi, & Sapiro, 2000).

3.2 Phase Two: Training and Testing

Dataset, Applied ML and DL

Algorithms and Evaluation Process

3.2.1 Training and Testing Dataset

To train the data according to our study, we used the

Python TensorFlow package and divided the data into

two groups, a training group and a test group, in ratios

of 8:2, according to the table 2 explanation for the

training and testing processes for the data sets.

Table 2: Traning and Testing Dataset.

Tree Diseases Trainin

Testin

Corn

Common Rust 1907 477

Vercospora Leaf

1642 410

Northern Leaf Blight 1908 477

Health

1859 465

Potato

Earl

Bli

ht 1939 485

Late Bli

ht 1939 485

Health

1824 456

Tomato

Bacterial Spot 1702 425

Septoria Leaf Spot 1745 436

Target Spot 1827 457

Health

1926 481

3.2.2 Applied ML and DL Algorithms

In this section, artificial intelligence algorithms,

specifically the VGG19 algorithm and the Support

Vector Machine (SVM) algorithm, were presented.

They are used for the detection and classification of

plant diseases. Additionally, we incorporated SIFT

and the Gabor filter in our work to enhance the results

obtained with the Support Vector Machine algorithm.

The outcomes demonstrated significant

improvements compared to our previous studies (Al-

Shalout et al., 2023).

VGG19

The primary contribution of the VGG network lies in

its focus on augmenting the depth of the

convolutional neural network to improve accuracy

(Yin, Wortman Vaughan, & Wallach, 2019). This is

accomplished by replacing a single 5×5 convolution

layer with two layers of size 3×3, and substituting one

7×7 convolutional layer with three 3×3 convolutional

layers. This structural adjustment not only increases

the network's depth but also minimizes the number of

parameters needed for the model (Qi, 2024). The

architecture of the VGG19 network is depicted in

Figure 3.

Figure 3: VGG 19 Network Structure (Qi, 2024).

SVM Classifier

Support Vector Machine (SVM) (Vapnik, 1998) is a

supervised machine learning algorithm used for both

classification and regression tasks. It's particularly

effective in scenarios where the data is not easily

separable through linear boundaries. SVM works by

finding the optimal hyperplane that maximally

separates different classes in the feature space (see

Figure 4).

Figure 4: Principle of Support Vector Machine; two-class

hyper-plane example (Elleuch, Maalej, & Kherallah, 2016).

A Leaf Disease Detection Using Machine Learning and Deep Learning: Comparative Study

943

SIFT

Scale-Invariant Feature Transform (SIFT) is a

computer vision algorithm that was introduced by

David G. Lowe in 1999 (Lowe, 1999). It is widely

used for object recognition, image stitching, and other

applications in computer vision and image

processing. SIFT is particularly powerful because it

is invariant to changes in scale, rotation, and

illumination, making it robust in various real-world

scenarios.

Gabor Filters

Gabor filters have become prominent in the domain

of pattern recognition. The primary focus of Gabor

filters is their ability to remain invariant to translation,

rotation, scale, and illumination variations. We

directly derive features from gray-level images using

Gabor filters, allowing us to extract pertinent

information in both spatial and frequency domains

(Daugman, 1985; Jain & Farrokhnia, 1991).

3.2.3 Evaluation Process

In this paper, we used many metrics to evaluate the

effectiveness of the proposed approach, and we

adopted Accuracy, loss function, and Recall, in

addition to F1 score, Precision and finally Confusion

Matrix.

- Accuracy: ACC= (TP+TN)/(TP+FP+FN+TN)

- Precision: Precision = TP/(FP+TP)

- Recall: Recall =TP/(TP+FN)

- F1 score: F1 Score= 2*(Recall*Precision) / (Recall

+ Precision)

- Loss Function:

Mean Squared Error =(y



− y



)







n 

4 EXPERIMENTAL RESULTS

AND DISCUSSION

In this study, we conducted many experiments to

verify the proposed algorithms for predicting diseases

of target plants, namely potato, tomato, and corn.

First of all, we performed the operations via the VGG

19 algorithm. The study was carried out on the entire

number of images, i.e. 25272 real images. The results

are shown in Table 3.

The results showed that the algorithm works

excellently, as the accuracy rate in corn plants

reached 97%, while in potatoes it reached 96%, and

finally in tomatoes it reached 95%, which is

considered an excellent result in relation to the

number of data.

Table 3: Experimental results using VGG19 algorithm.

Tree Recall Recall F1 Accurac

Corn

Common

Rust

0.99 0.92

0.97

Vercospor

Leaf S

0.79 1.00

Northern

Leaf

Blight

0.66 0.98

Health

0.89 0.29

Potato

Early

Bli

0.78 0.96

0.96

Late

Bli

0.99 0.62

Health

0.87 0.99

Tomato

Bacterial

0.92 0.95

0.95

Septoria

Leaf Spot

1.00 0.88

Target

0.98 0.79

Health

0.29 0.44

Non-trainable parameters refer to the number of

parameters in a neural network model that are not

updated or learned during the training process. In this

study, we mention that there are 20024384 non-

trainable parameters in our VGG19 model (see Table

4).

Table 4: Parameters of VGG19 Model.

Layer (type) Output Shape Param #

vgg19 (Functional) (None,7,7,512) 20 024 384

flatten (Flatten) (None,25088) 0

dense

(

Dense

)

(

None , 4

)

100 356

Total params: 20 124 740

Trainable

arams: 100 356

Non-trainable params: 20 024 384

Figure 5: Performance of our proposed Model (VGG19).

Figures 5 and 6 illustrate the effectiveness of the

VGG19 algorithm in detecting diseases in potato

plants. The figure visually represents the significant

enhancement in results, with 35 Epoch of successful

ICAART 2025 - 17th International Conference on Agents and Artiﬁcial Intelligence

944

disease detection in the study, achieving a remarkable

accuracy rate of 96% for this plant.

Figure 6: Trainig and validation Loss function (VGG19).

The second experiment using the SVM, where the

feature was extracted through SIFT and Gabor Filter,

achieved excellent results, more than expected, and

with the same number of images, we obtained an

accuracy rate that reached 96% in the tomato plant,

98% in the corn plant, and finally 97% in potato

plants. The algorithm clearly excelled in detecting

diseases that affect plants, and the Table 5 shows the

results that appeared in the study and the experiments.

Table 5: Experimental results for SVM algorithm.

Tree Recall Recall F1 Accurac

Corn

Common Rust 0.94 0.95

0.98

Vercospora

0.98 0.99

Northern Leaf

Blight

0.99 0.98

Health

0.91 0.93

Potato

Earl

Bli

ht 1.00 0.88

0.97

Late Bli

ht 0.91 0.92

Health

0.89 0.98

Tomato

Bacterial Spot 0.98 0.95

0.96

Septoria Leaf

Spot

0.96 0.98

Target Spot 0..87 0.93

Health

0.92 0.94

Table 6 describes the model summary for SVM

algorithm used in experiments.

Table 6: Summary details for SVM Model.

Train Data - Train Labels Test Data - Test

Labels

features shape (5702, 25088) (1426, 25088)

labels sha

(

5702

)

(

1426

)

latted train and test data

Train data (5702, 25088)

test data (1426, 25088)

Train labels (5702)

test labels (1426)

Figures 7 and 8 depict the model's performance in

detecting plant diseases using an SVM algorithm

specifically for potato plants. The figure 7 visually

demonstrates the improvement in results in terms of

the number of epochs, with the study yielding 18

epochs. The accuracy percentage for this plant

reached 97%, indicating a slight improvement

compared to the previous algorithm, where it was

96% for potato plants.

Figure 7: Performance of our proposed SVM Model.

Figure 8: Training and validation Loss function (SVM).

Figures 9 and 10 display the confusion matrix

regarding tomato plants, with an accuracy rate of

96%. Upon integrating the Gabor filter, the accuracy

improved, reaching 97%. This visualization

highlights the efficacy of the proposed system in

categorizing tomato plants across four distinct

classes.

Figure 9: Confusion matrix about VGG model.

A Leaf Disease Detection Using Machine Learning and Deep Learning: Comparative Study

945

Figure 10: Confusion matrix about SVM model.

The results were notably favorable, particularly with

the Support Vector Machine algorithm achieving an

impressive accuracy rate of 98% on our dataset across

multiple classes in a single stage. This high accuracy

rate is commendable. Meanwhile, the VGG19

algorithm achieved an accuracy rate of 97% on the

same dataset, representing a notable improvement

compared to earlier studies using the VGG16

algorithm, where the accuracy was 89% (Shijie et al.,

2017).

Several researchers have utilized diverse models,

incorporating CNN algorithms along with Transfer

Learning, AlexNet, and ResNet. Their results have

shown a range of accuracies, varying from 83.46% to

as high as 95.71%. These findings highlight the

significance of numerous studies focused on

identifying plant diseases, underscoring the crucial

role of plants in human life (see Table 7).

Table 7: Comparative results.

Id Dataset Al

orithms Accurac

(Natarajan

et al.,

2020)

1090 real

images

ResNet,

AlexNet, and

Squeeze Net

95.71%

(Shijie et

al., 2017)

7040

images

CNN model

with transfer

learning and

VGG16

89%

(Arora &

Agrawal,

2020

)

12332

images

LeNet5 83.46%

CNN 91.25%

(Mohanty

et al.,

2016)

54306

images

CNN 99.35%

Current

work

25272

ima

SVM (Gabor

filters & SIFT

)

98%

25272

ima

VGG19 97%

5 CONCLUSION

In this study, we targeted three types of plants in the

study and a number of diseases that affect them,

which were leaf spot, northern leaf blight, early and

late blight, in addition to rot and bacterial rust on the

plants.

The results showed that the proposed methods,

which were added manually to improve the images

using the Gabor filter in addition to Sift, achieved a

high accuracy rate in detecting diseases that affect

plants, reaching 98%, which is a good thing,

especially since the number of data used in the study

is large, Furthermore, an additional step in the future

involves employing data augmentation to augment

the number of images, thereby enhancing the results

further.

5.1 Future Work

Based on the findings, future work could focus on

enhancing disease detection algorithms to improve

their efficiency and accuracy in dealing with diverse

and large datasets. This could be achieved by

exploring advanced techniques such as transformers

or incorporating ensemble learning methods to

improve model performance. Furthermore, the study

could be expanded to include a wider range of plant

diseases and species, with a focus on early detection

at the initial stages before visible symptoms appear,

enabling faster and more effective interventions. The

integration of advanced technologies such as edge

computing and the Internet of Things could facilitate

real-time data collection and analysis, contributing to

the development of smart systems that leverage AI for

automated disease detection. These systems could be

designed for practical field use through mobile

applications or dedicated devices, making them

accessible to farmers. In addition, these systems could

support sustainable agriculture by reducing reliance

on chemical pesticides and achieving more

sustainable improvements in agricultural

productivity. Finally, it is recommended to enhance

collaboration with experts in plant science and

agriculture to develop comprehensive and integrated

solutions, leveraging modern technologies to create a

clear improvement in the agricultural sector.

ICAART 2025 - 17th International Conference on Agents and Artiﬁcial Intelligence

946

REFERENCES

Al-Shalout, M., Elleuch, M., & Douik, A. (2023). Detection

Plant Diseases Using Deep Learning Algorithms. Paper

presented at the 2023 International Conference on

Cyberworlds (CW).

Albawi, S., Mohammed, T., & Al-Zawi, S. Understanding

of a convolutional neural network. In2017 international

conference on engineering and technology (ICET) 2017

Aug 21 (pp. 1-6): IEEE.

Alzahrani, M. S., & Alsaade, F. W. (2023). Transform and

deep learning algorithms for the early detection and

recognition of tomato leaf disease. Agronomy, 13(5),

1184.

Arora, J., & Agrawal, U. (2020). Classification of Maize

leaf diseases from healthy leaves using Deep Forest.

Journal of Artificial Intelligence and Systems, 2(1), 14-

26.

Daugman, J. G. (1985). Uncertainty relation for resolution

in space, spatial frequency, and orientation optimized

by two-dimensional visual cortical filters. JOSA A,

2(7), 1160-1169.

Elleuch, M., Maalej, R., & Kherallah, M. (2016). A new

design based-SVM of the CNN classifier architecture

with dropout for offline Arabic handwritten

recognition. Procedia Computer Science, 80, 1712-

1723.

Goralski, M. A., & Tan, T. K. (2020). Artificial intelligence

and sustainable development. The International

Journal of Management Education, 18(1), 100330.

Jain, A. K., & Farrokhnia, F. (1991). Unsupervised texture

segmentation using Gabor filters. Pattern recognition,

24(12), 1167-1186.

Laranjeira, T., Costa, A., Faria-Silva, C., Ribeiro, D., de

Oliveira, J. M. P. F., Simões, S., & Ascenso, A. (2022).

Sustainable valorization of tomato by-products to

obtain bioactive compounds: Their potential in

inflammation and cancer management. Molecules,

27(5), 1701.

Lazzeri, F., Bruno, G., Nijhof, J., Giorgetti, A., & Castoldi,

P. (2015). Efficient label encoding in segment-routing

enabled optical networks. Paper presented at the 2015

International Conference on Optical Network Design

and Modeling (ONDM).

Lowe, D. G. (1999). Object recognition from local scale-

invariant features. Paper presented at the Proceedings

of the seventh IEEE international conference on

computer vision.

Mohanty, S. P., Hughes, D. P., & Salathé, M. (2016). Using

deep learning for image-based plant disease detection.

Frontiers in Plant Science, 7, 1419.

Natarajan, V. A., Babitha, M. M., & Kumar, M. S. (2020).

Detection of disease in tomato plant using Deep

Learning Techniques. International Journal of Modern

Agriculture, 9(4), 525-540.

Qi, J. (2024). Efficiency study of VGG networks in

autonomous driving tasks. Paper presented at the

Fourth International Conference on Computer Vision

and Data Mining (ICCVDM 2023).

Reis, H. C., & Turk, V. (2024). Integrated deep learning and

ensemble learning model for deep feature-based wheat

disease detection. Microchemical Journal, 197,

109790.

Shijie, J., Peiyi, J., & Siping, H. (2017). Automatic

detection of tomato diseases and pests based on leaf

images. Paper presented at the 2017 Chinese

automation congress (CAC).

Shruthi, U., Nagaveni, V., & Raghavendra, B. (2019).

review on machine learning classification techniques

for plant disease detection. Paper presented at the 2019

5th International conference on advanced computing &

communication systems (ICACCS).

Singh, D., & Singh, B. (2020). Investigating the impact of

data normalization on classification performance.

Applied Soft Computing, 97, 105524.

Song, Z., Fu, L., Wu, J., Liu, Z., Li, R., & Cui, Y. (2019).

Kiwifruit detection in field images using Faster R-CNN

with VGG16. IFAC-PapersOnLine, 52(30), 76-81.

Vapnik, V. (1998). Statistical learning theory. John Wiley

& Sons google schola, 2, 831-842.

Weinberger, M. J., Seroussi, G., & Sapiro, G. (2000). The

LOCO-I lossless image compression algorithm:

Principles and standardization into JPEG-LS. IEEE

Transactions on Image processing, 9(8), 1309-1324.

Yin, M., Wortman Vaughan, J., & Wallach, H. (2019).

Understanding the effect of accuracy on trust in

machine learning models. Paper presented at the

Proceedings of the 2019 chi conference on human

factors in computing systems.

A Leaf Disease Detection Using Machine Learning and Deep Learning: Comparative Study

947